Anthropic, the company behind the Claude AI assistant, is putting together a team to design its own computer chips. Reuters first reported in April 2026 that the company was exploring the idea; by August, Anthropic had posted job listings for a chip design engineer role based in San Francisco and New York TechCrunch. The company's careers page also lists a Hardware Security Engineer position, suggesting the effort covers chip security as well as performance.

Anthropic said it plans to design hardware and AI models together to make its technology faster and more efficient TechCrunch. The Information reported that Anthropic has been talking to Samsung as a potential partner to actually manufacture the chips, though no deal has been confirmed TechCrunch.

The push comes as demand for Claude grows and AI companies compete to secure enough computing power. Reuters reported in April that a shortage of AI computing hardware is driving the effort CNBC. At that time, Anthropic's plans were still preliminary. The August hiring push shows the company has moved beyond the thinking stage.

Today, Anthropic runs Claude on a mix of chips from different companies. It uses AWS Trainium, Google TPUs, and NVIDIA GPUs, choosing the best chip for each task Anthropic. In April 2026, Anthropic expanded its partnership with Google and Broadcom. Weeks later, it extended its deal with Amazon for up to five years of computing capacity Anthropic. Anthropic also plans to use as many as one million Google Cloud TPUs Anthropic. On the AWS side, Anthropic's engineers work closely with Amazon's own chip design team to get the most out of Trainium hardware Anthropic. The company has signed deals with AWS, Google, Nvidia, and AMD for AI computing hardware access TechCrunch.

Anthropic is not the first AI company to go down this road. Google has used its own TPU chips for years. Meta is developing its own AI chips. In June 2026, OpenAI unveiled a chip called Jalapeño, built by Broadcom and designed specifically for running AI models TechCrunch. The pattern is the same across these companies: as AI models get bigger and more people use them, buying off-the-shelf chips becomes too expensive and too limiting. Custom chips, designed for a company's own specific needs, offer a way around that.

What sets Anthropic apart is how many chip partners it already has. Few AI companies work as closely with AWS, Google Cloud, Broadcom, Nvidia, and AMD all at the same time. Designing its own chips is less about replacing those partners and more about gaining control over details that those partners' chips were not built to handle — things like how fast the chip responds, how quickly it moves data in and out of memory, or how efficiently it runs Anthropic's specific AI workloads.

The job listing for a reinforcement learning-focused chip design engineer is worth noting. Reinforcement learning is a type of AI that learns through trial and error. Google has already used this approach to arrange the layout of components on its TPU chips, producing designs that beat what human engineers could create. Anthropic hiring for this specific skill suggests the company wants to use its own AI models to help design its chips, rather than simply handing a list of requirements to an outside design firm.

Anthropic's careers page listed 398 open jobs in AI research and engineering as of the most recent access, with the custom chip roles sitting inside that broader hiring surge. The scale of recruiting signals that Anthropic is building out its hardware capabilities alongside its AI models, treating the design of chips and software as one connected effort rather than simply buying what it needs.

Several questions remain unanswered: whether Anthropic will actually complete a chip design, which factory will manufacture it, and whether Samsung will be the partner that makes it happen. The company has not committed to a specific chip architecture or product. What is clear is that four months ago Anthropic was just considering custom chips, and today it is hiring the people to build them. For a company whose products depend entirely on the economics of running AI models, that direction makes sense on its own terms.